It is difficult for the bioactive peptides whose sequence lengths are different to establish QSAR （Quantitative structure-activity relationship） model. Although it could be solved by ACCs （Auto and cross auto covariances）， its estimation ability is not satisfactory. Thus， a new method named TTPN （Two-terminal position numbering） was proposed. The principle of TTPN method is to find the shortest length peptide and standardize its number of amino acids， the information of same number amino acid residues of the C-terminus and N-terminus was extracted from database. Then the descriptive variable matrix X and the active data matrix Y were established for QSAR research. In order to verify the effectiveness and application scope of TTPN method， many peptides involve the BT （Bitter peptides）， ACE （Angiotensin I-converting enzyme inhibitor） and antioxidant peptides measured by ORAC （Oxygen radical absorption capacity） were collected. All the results of QSAR showed that TTPN method was better than ACCs to describe the sequence of peptide （R2 and Q2 for TTPN and ACCs are 0.724， 0.329； 0.599， 0.038 in ACE databases respectively）. Furthermore， TTPN could also describe sequences in which information was partially or fully described （R2 and Q2 for 2-14 and 7-14 databases are 0.717， 0.693 and 0.922， 0.753 respectively）. In conclusion， TTPN method not only provides an effective method for the establishment of QSAR model of peptide with distinct length， but also has a better analysis and prediction ability as well as a wider range of applications compared with ACCs method. The explanation of active peptide structure characteristics obtained from the model compared with ACCs method are more precious.